The Things We Do Best
Artificial Intelligence
Our goal and mission are the creation of intelligent systems that use machine learning to improve their performance without human intervention. Over the years, nSpace Analytics data scientists have invented/co-invented many automotive AI processes for which patent applications have been awarded in the U.S., Germany, and China. The first patent on which we led the research was issued to General Motors in 2010 for an intelligent cruise control mechanism that utilized a novel Markov Chain State Model to sense a vehicle’s driving environment and driver input to prevent the resumption of an earlier set high speed setting that might be inappropriate for the current driving situation. This was an exciting and novel application of artificial intelligence! Many additional patents have followed.
Big Data Management
We support our machine learning and predictive modelling analysis by creating and tasking petabyte-level databases that reside in high-performance big data clusters, using such technologies as Hadoop. In the automotive world, data is collected from test vehicles equipped with CANbus in-vehicle sensor networks. These networks typically capture hundreds of vehicle signals on a real-time basis, at up to 10 hz data capture rates. On a daily basis, gigabyte-scale files are transmitted through the cloud to the clusters. To analyze this data, nSpace data scientists examine the database schema to identify signals of interest. Hive queries are created to extract and decode the desired signals. Vehicle signal data is often augmented with data from online map databases.
Predictive Analytics & Modelling
nSpace Analytics conducts on-going research to develop predictive models that can be deployed in a variety of industrial settings. A current internal research project involves the development of an Adaptive Engine Stop/Start Algorithm that predicts when an automobile is expected to soon come to a stop. In such cases, the vehicle’s internal combustion is immediately shut down while the vehicle is still in motion and approaching a stop. The engine is restarted when the driver depresses the accelerator pedal. This innovation improves fuel economy and reduces carbon emissions further than the current engine stop/start technology which delays engine shutdown until the vehicle has stopped. Using a Classification and Regression Decision Tree machine learning model, preliminary test results show a short-range stop prediction accuracy of better than 94%.